1 / 30

Irina Gladkova 1,2,3 , Yury Kihai 1 ,2 , Alexander Ignatov 1 ,

Irina Gladkova 1,2,3 , Yury Kihai 1 ,2 , Alexander Ignatov 1 , Fazlul Shahriar 3,4 , Boris Petrenko 1,2 1 NOAA/NESDIS/STAR, 2 GST, Inc., 3 City College of New York, NOAA/CREST, 4 Graduate Center of CUNY. SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014.

rufin
Download Presentation

Irina Gladkova 1,2,3 , Yury Kihai 1 ,2 , Alexander Ignatov 1 ,

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Irina Gladkova1,2,3, Yury Kihai1,2, Alexander Ignatov1, Fazlul Shahriar3,4, Boris Petrenko1,2 1NOAA/NESDIS/STAR, 2GST, Inc., 3City College of New York, NOAA/CREST, 4 Graduate Center of CUNY. SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014 Exploring Pattern Recognition Enhancements to ACSPOClear-Sky Mask for VIIRS: Potential and Limitations.

  2. Motivation • ACSPO Clear-Sky Mask (ACMS) employs comparisons of retrieved SST with L4 analyses, reflectance threshold tests and spatial uniformity tests. • ACSM performs well on a global scale but tends to over-screen some highly dynamic areas (e.g., with strong currents, cold upwellings, eddies) as well as the coastal zones. • These deficiencies cannot be completely eliminated by simple thresholds adjustment within ACSM without triggering massive cloud leakages. • Visual analysis of SST field easily discriminates cloud leakages from cold SST anomalies SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  3. VIIRS SST SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  4. ACSPO Clear Sky Mask SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  5. Types of misclassifications • Typical clear sky ocean regions misclassified by the ACSM : • contiguous, • with well-defined boundaries, • typically located in the vicinity of ocean thermal fronts. • Existing image processing techniques: • Segmentation; • Morphological Procedures: erosion and dilation; • Thermal Front Detection. SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  6. Human Perspective • Humaneye does not perceive absolute pixel values (i.e., SST values) • It relies instead on local contrasts and ratios, which more directly correlate with gradients in an image. • Difference between ocean and cloud patterns should be more pronounced in the SST gradient magnitude domain. SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  7. Magnitude Gradient Gradient magnitudes viewed as a terrain look like sharp ridges towering over flat valleys. SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  8. Algorithm Step 1: Identify Search Domain Step 2: Determine SST gradient ridges Step 3:Determine spatially connected cold SST regions Step 4: Discard SST segments found in Step 3 that do not border the ridges found in Step 2 Step 5: Statistical Test SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  9. Search Space SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  10. Gradient Ridges SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  11. Segmentation SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  12. Adjacency SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  13. Statistical Test SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  14. Clear Sky Regions SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  15. SST Gradient Ridges • Existing Image Processing Tools: • Thermal Front Detection • Edge Detection • Gradient Ridges and Valleys SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  16. Border Stability SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  17. SST Gradient Ridges SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  18. Watershed Segmentation • Segmentation/Clustering is a well studied field • Many ways to perform segmentation • We use watershed type applied to ΔSST SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  19. Steps of Segmentation Segments obtained via iterative procedure: Iter0:Initial segments Iter k:Lower the threshold level Find new “catchment basins” Re-label in case of split SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  20. Practical Considerations • Bow-Ties • Striping • Missing Reference SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  21. Potentials and Limitations • Considered 2 sets of VIIRS data: • 48 hand picked and cropped regions with typical clear sky misclassification • 144 granules representing 1 day global observations Results were visually inspected and analyzed; Success rate is promising but more work is needed. SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  22. South Africa, 02/17/13 (night) SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  23. South Africa, 02/17/13 (night) SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  24. Gulf Stream, 05/10/13 (day) SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  25. Gulf Stream, 05/10/13 (day) SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  26. Uruguay, 05/05/13 (night) SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  27. Uruguay, 05/05/13 (night) SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  28. Pamlico Sound, 02/16/13 (night) SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  29. Pamlico Sound, 02/16/13 (night) SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

  30. Conclusion • A supplemental algorithm to the current ACSPO Clear-Sky Mask based on pattern recognition is being explored. • Our preliminary analyses suggest that some of the limitations inherent to the current ACSM may be alleviatedand SST coverage improved. • The improvements are mostly noticeable in the areas interesting to ACSPO users, including dynamic areas of the ocean and coastal zones. • Future work will include tuning the algorithm, with emphasis on resolving the remaining cloud leakages. SPIE, Ocean Sensing and Monitoring VI, Baltimore, 5-9 May 2014

More Related